[V3dot10] making and recording changes

[David T. Lewis]

On Thu, Jan 18, 2007 at 06:44:56PM +0100, stephane ducasse wrote:
> 
> On 18 janv. 07, at 18:26, Ralph Johnson wrote:
> 
> >On 1/18/07, stephane ducasse <stephane.ducasse at free.fr> wrote:
> >>It seems to me that I already said that. But ask and I will repeat.
> >
> >We did talk, but somehow I never realized that.
> >
> >>Ralph this is what we DID for 3.9.
> >
> >So, why did people complain that you did not have an update stream?
> 
> 
> because it was slow (we did not use MCConfig) and sometimes we could not
> really chain well so we had to do changes by hand. So it was  
> necessary to load from
> a given image. But this was not the process the problem just the  
> difficulties we encountered.
> 
> Because people like to complain.

<gripe>
I do not usually like to complain, but I am going to do so now. There
was no update stream for 3.9. I had to download complete images in
order to do testing, which means that I could not easily maintain an
image to use help for testing 3.9 updates. It required a lot of time
and effort just to download the latest image, move things from that
last test image into the new one, add packages that I personally
maintain, and so forth. I did try to help, but usually it was too
much work and not enough fun, so I did not put much time into it.
Neither did anyone else, so you and Marcus ended up doing a whole
lot of work without much support from other people.

A real update stream is much better, because it permits someone like
me to maintain a "bleeding edge" testing image and keep it up to date
with the release stream. Any time an interesting problem shows up on
Mantis, it is possible to work on it in the testing image and try
out proposed fixes. That's a nice way to spend an hour or two over
a cup of coffee.

I don't care if the update stream uses change sets or something
else, but I would really like to see an update stream again so that
hundreds or maybe thousands of Squeak users can stay up to date and
do their part to help with testing.

I would also add that a real update stream provides a clear, documented
record of what went into the image, when it happened, who it came
from, and why it was done. The change sets identify their authors,
and method changes are properly stamped with (real) author initials
and time stamps. I cannot think of any acceptable excuse for losing
all of this information, yet somehow that is exactly what we have
done with our "improved" processes.

OK, I'm done complaining.
</gripe>

Dave